This study was preregistered on PROSPERO (International Prospective Register of Systematic Reviews; CRD42025641383) and has been reported according to the RAMESES (Realist and Meta-Narrative Evidence Syntheses: Evolving Standards) [14].

The glossary of terms used to describe the steps undertaken to carry out this review has been presented in Table 1.

The following 5 steps were taken to conduct the realist review.

Initial program theories were cocreated in workshops with the stakeholders of a pilot project that aimed to explore the efficacy of a smart medication device called the “Medybox” in improving medication adherence among domiciliary care recipients and hospital discharge patients in a region in Northern England, United Kingdom. In the workshops, the stakeholders hypothetically discussed the features of the Medybox (a modified electronic pillbox) as an example of a digital medication adherence intervention, which informed the realist review that explored the program theories of all types of digital medication adherence interventions to reduce unintentional medication nonadherence. These stakeholders were presented with a summary of the features of the device and its implementation process, summarized as follows:

Six cocreation workshops of approximately 2 hours each in 2 workshop rounds (3 workshops and 12 participants in each round) were conducted to generate ideas for program theories for the implementation of this particular device. Stakeholders included members of the regional National Health Service integrated care board (ICB) who were former health care professionals (nurses) but currently working on a digitizing adult social care program, device providers who were implementing the intervention, ICB pharmacists, managers of domiciliary care organizations, individuals receiving care for their medication intake and their informal or family caregivers. Most of the stakeholders were women (10/12, 83%) and 83% (10/12) were White British (consistent with the ethnicity of the population in the local area).

In keeping with qualitative methods, a small number of appropriate stakeholders were recruited purposively via word of mouth and contacts of the research team, the ICB, and the device providers. Stakeholders were “experts by experience” as they had lived experiences of receiving or providing care to individuals requiring help with medication adherence. ICB professionals were also knowledgeable about possible contextual factors that could lead to unintended outcomes of the program [13,18]. The same group of professionals and care recipients were invited to participate in the second round of workshops. Additionally, participants were recruited if a certain group of stakeholders was underrepresented in the first round. One of the device providers involved in the implementation of the program dropped out after the first round. Therefore, the researchers recruited an ICB pharmacist in the second round to maintain numbers and improve diversity of the stakeholders involved in the workshops (see Saha et al [19] for detailed information about cocreation workshops). The workshops did not include academic researchers as stakeholders.

The discussions were facilitated by the research team, who provided participants with prompts regarding potential contexts (eg, technological literacy of individuals), mechanisms (medication intake reminders), and outcomes (eg, medication adherence) that were derived from the emerging findings of this review. Specific parts of the workshop discussions were audio- or audio-and-video-recorded, and the research team made notes of the workshop discussions. The context-mechanism-outcomes (CMOs) derived from the first round of workshops were presented to the participants in the second round of workshops to further refine them. A list of initial ideas of CMOs derived from workshop discussions was triangulated with the findings from the realist review. This was an iterative process in which the workshop discussions informed the realist review and vice versa. Further details about the device and the prompts used to facilitate workshop discussions are presented in Multimedia Appendix 1. This method of conducting cocreation workshops has been used to explore initial program theories by other realist reviews [20].

Searches were conducted on the following databases from January 2002 till December 2024: MEDLINE, CINAHL, PsycInfo, Web of Science, Scopus, Embase, and gray literature databases called Overton and Policy Commons. Hand searching of reference lists of reviews was conducted to identify any additional studies. The research team started searching from January 2002, as technologies such as smart pillboxes (devices that store and dispense scheduled medication doses at correct times, provide medication intake reminders, and monitor medication adherence by recording when the user opens the pillbox), for example, Pivotell, were first used extensively in the United Kingdom in that year [21]. Thus, these search dates increased the likelihood of inclusion of studies related to recent digital technologies designed to improve medication adherence such as smart pillboxes, automated interactive voice recognition, and mobile apps [22]. The search terms were decided based on the population (eg, multiple chronic conditions), intervention (eg, digital intervention), and outcomes (eg, medication adherence) explored in this review. The full search terms are presented in Multimedia Appendix 2.

The papers were assessed for eligibility based on the inclusion and exclusion criteria presented in Table 2.

Titles and abstracts were screened by the first author using the web-based systematic review platform Rayyan [23] and around 10% (200/2152) of the papers were also screened by the other members of the research team for inclusion. Any disagreements were resolved through discussion.

A data extraction sheet was used to document details regarding authors, year of publication, study title, document type (eg, systematic review and pre-post intervention study), participants, and rigor, richness, and relevance scores. The data extraction sheet is presented in Multimedia Appendix 3 [3,9-11,29-107]. The rigor of the studies was judged based on the credibility and trustworthiness of the methodology, for example, sample size, data collection, and analysis methods [13]. The rigor of gray literature, opinion pieces, and book chapters was assessed based on significance, objectivity, coverage of the topic, authority, and accuracy [28]. Relevance was judged by assessing the extent to which papers presented information related to the topic of the review and provided evidence for program theory development. Richness was judged by assessing the extent to which papers provided details regarding mechanisms, contexts, and outcomes of the interventions [12,13]. The scoring criteria used in the realist review by Waldron et al [12] were used to assess the rigor, relevance, and richness. Studies were scored on a scale of 0 (very poor) to 3 (very good) in terms of relevance and rigor, and 0 (nothing of interest or not focused on design, implementation, or use) to 4 (much valuable data) in terms of richness. This helped in understanding the extent to which each study contributed to program theory development and the extent to which each program theory was supported by relevant and rigorous evidence [12,13].

The lead author manually annotated the papers for program theory ideas with the help of NVIVO (Lumivero, version 14) software. Data from the literature and workshops were analyzed separately. Patterns of context, mechanisms, and outcomes from the studies included were explored by using inductive and deductive logic, and relevant text to support each program theory was documented. The recordings of the workshop discussions were transcribed verbatim. The lead author reviewed the transcripts and the participant and facilitator notes of the workshop discussions and coded the patterns of contexts, mechanisms, and outcomes. Anonymized participant quotes and participant and facilitator notes supporting each CMO were documented.

The CMOs from the realist review were then triangulated with the CMOs from the cocreation workshops by noting the extent to which the CMOs from the workshops corroborated or not with the CMOs derived from the literature. After that, a final list of program theories in different theory areas and the number of CMOs from the literature, workshops, or both were noted. After the initial CMOs were created, an iterative process of revising and refining was conducted through discussion with the research team.

This study received ethics approval from the Clinical Research Sponsorship ethics committee at Lancaster University and the Health Research Authority and Health and Care Research Wales ethics committee on February 4, 2025 (ethics board: social care; reference number: 24/IEC08/0037).

Eighty-three studies were included in this review. Figure 1 depicts the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) diagram with details about the search and screening process.

Eighty-three papers, including peer-reviewed research papers (n=70) and gray literature (n=13), were included in the review. Of the research papers, 44 were narrative, systematic, or scoping reviews or meta-analyses, and 3 papers were expert opinion pieces. Of the 23 primary studies included, the largest number were conducted in the United States (n=9), with studies also conducted in Canada (n=4), Spain (n=2), the United Kingdom (n=2), China (n=2), Australia (n=1), and Switzerland (n=1). The country in which the study was conducted was not reported by 2 studies [29,30]. The primary studies included 17 intervention studies [3,10,29-44]. The rest were qualitative or quantitative studies that explored participants’ views (where participants were individuals diagnosed with chronic conditions, informal caregivers of patients diagnosed with chronic conditions, and health care professionals) regarding the components of digital medication adherence interventions [43-48].

Most of the studies were assigned a score of 3 in the “rigor” domain. However, approximately 60% (50/83) of the studies were assigned a score of 1 or 2 in the richness domain due to lack of detail regarding contexts and mechanisms related to the outcomes of the interventions. This was primarily because these studies did not report the content and delivery of the interventions in sufficient detail. Approximately 24% (20/83) of the studies were assigned a score of 1 in the “relevance” domain due to lack of data for program theory development. The authors triangulated data from studies that had low “rigor” scores with the data from studies that had higher scores while developing program theories [13,108]. The studies that had low “relevance” and “richness” scores were included in the review if they provided enough information for program theory development and if the data provided by them could be supported by the studies that were assigned high scores in these domains. The authors also transparently reported if a certain program theory was not supported adequately by rigorous, relevant, and rich data [108]. Details about the studies included in the review are presented in Multimedia Appendix 3.

Thirty-five CMOs grouped in 7 theory areas were derived from the review, cocreation workshops, or both. A narrative summary of the CMOs derived from the literature, cocreation workshops, or both in each theory area is presented in the following section. The CMOs are presented in Multimedia Appendix 4 [32,35,38,40-42,45,47,48,52,56,74,79,87,97,100,106], along with specific quotes to support them that were derived from workshop discussion recordings and/or studies in this review.

Overview

This theory area includes the CMOs related to ways by which digital interventions can improve collection of prescription medications on time. These CMOs were derived from the literature only. The overarching program theory presented in Figure 4 is based on the CMOs of theory 3 presented in Multimedia Appendix 4.

Studies have highlighted the importance of enabling easier access to prescription medications to prevent primary nonadherence. Text message reminders or reminders from mobile apps to order prescription medications on time and outlining the exact process of ordering prescriptions and collecting them can improve medication adherence. This would be helpful for individuals who forget to order prescriptions or who are unable to understand the process of ordering repeated prescriptions. Additionally, in the context in which frequent medication reviews are not required, providing repeat prescriptions for longer periods of time can improve medication adherence as it would reduce the frequency of ordering prescriptions (CMO 13) [48,68,79,80,99].

Mash et al [100] suggested that collecting prescription medications on time could be particularly challenging for individuals living in low- and middle-income countries (LMICs) due to limited human resources and poor infrastructure in primary health care services, resulting in long waiting periods to collect prescriptions. Thus, this study [100] suggested the installation of pharmacy dispensing units which allow remote dispensing of medications (medication dispensing can be requested online by patients) and storage of medicines in password-protected lockers in convenient areas. This can allow patients to collect their medicines at a convenient time and help in improving medication adherence (CMO 14). However, installation of pharmacy dispensing units requires high start-up costs. Additionally, the long-term efficacy, feasibility, and acceptability of these technologies are yet to be determined, as these have been piloted on a very small scale only in South Africa [77,100].

This theory area includes the CMOs related to how the smart medication device that was provided to domiciliary care recipients and hospital discharge patients in the United Kingdom could help in improving the medication adherence of patients with high anxiety, low self-efficacy, and low social support. The CMOs in this section were derived from the workshop discussions only. The overarching program theory presented in Figure 6 is based on the CMOs of theory 5 presented in Multimedia Appendix 4.

Domiciliary care recipients in the United Kingdom who have difficulties with medication adherence are generally visited by care providers to prompt them to take their medications. Therefore, participants discussed that if a domiciliary care recipient lives alone and has low social support, they may enjoy the company of care providers and look forward to their visits. These care recipients may feel isolated if the care providers’ visits are reduced when they self-manage their medicines. Thus, participants suggested that the device provider should assess the mental health and the potential for social isolation of these individuals and provide appropriate social support after the device is installed in their homes. This is especially important for individuals with a prior history of mental health diagnoses such as depression (CMO 20).

As mentioned in the Methods section, this device was provided to the recipients for only 6‐12 weeks, and during that time, the device was aimed at helping them to develop effective medication intake habits, after which the device was taken away. Participants mentioned that the removal of the device could cause anxiety among some individuals as they may become dependent on it for maintaining their medication adherence. Participants recommended that users could be allowed to keep the device that stores all the medicines safely, but the electronic reminder, which is used to provide medication intake reminders and monitor individuals’ adherence (the costly part of the device), could be taken away by the device provider. This could ensure that all the medicines are stored in the same place and strengthen individuals’ habits of taking medications from the same place, resulting in reduced anxiety and maintenance of medication adherence. However, some individuals with high anxiety may refuse to use the device at the outset if they are informed that it would be taken away after 6‐12 weeks. This could reduce the large-scale adoption and scalability of the intervention. The device provider also mentioned that they could suggest, following removal, alternative devices to maintain their medication adherence, for example, mobile apps. This could help individuals to try different types of technology to manage their medications and improve their comfort with technology (CMO 21).

Thirty-five CMOs in 7 theory areas were identified. CMOs within theory 1 demonstrated ways in which devices can help establish and maintain medication intake routines, such as by providing medication intake reminders. Theory 2 presented CMOs related to solving medication-related barriers to adherence, for example, self-monitoring of medication adherence to reduce double or missed doses. Theory 3 presented CMOs related to enabling collection of prescription medications on time, for example, by sending text message reminders to collect prescriptions on time. Theory 4 included CMOs related to improving the medication adherence of individuals with sensory and motor impairments and unpredictable health conditions, for example, encouraging individuals with visual impairments to use assistive devices with voice command features such as Alexa to receive medication intake reminders. Theory 5 presented CMOs related to customizing smart medication devices according to the needs of individuals with high anxiety and low social support and low medication self-efficacy, such as assessing the mental health of individuals who are socially isolated and providing them with appropriate social support. Theory 6 presented CMOs related to supporting individuals who are uncomfortable with technology, for example, providing training on how to use smart medication devices. Theory 7 presented CMOs related to policy and economic factors that were barriers to the large-scale implementation of smart medication devices, such as a lack of suitable incentives for care providers to refer domiciliary care recipients to use the device.

CMOs were derived from both the literature and workshops (14/35, 40%), with a proportion derived only from the literature (7/35, 20%) and a proportion only from the workshops (14/35, 40%). CMOs derived from workshops provided important information about implementing smart medication devices in the United Kingdom and in the context of domiciliary care and hospital discharge. CMO 14 (enabling easier access to prescriptions in LMICs) was not adequately supported by high-quality evidence, and further research is needed to support it. After assessing the “rigor” at a program theory level, the authors excluded a CMO that was initially developed regarding ingestible digital pills for monitoring and providing feedback on medication adherence because these have recently been discontinued due to low acceptability and adoption [114].

The results of the current realist review are consistent with those of a qualitative study [114] in which patients diagnosed with cardiovascular diseases acknowledged that medication intake reminders and reminders for keeping water ready to take with medications helped in maintaining effective medication intake habits. Similarly, the importance of tailoring the timing and frequency of the reminders according to individual preferences to prevent “alert fatigue” and annoyance related to receiving repeated reminders was also highlighted [115]. The importance of action planning related to medication intake by linking it with daily routine tasks such as mealtimes [116,117] was also supported by previous studies.

Previous studies have also found that self-monitoring and monitoring and feedback on adherence by health care providers could help in improving adherence, including reducing the possibility of missed doses [118,119]. Smart pillboxes or medication dispensers that allow monitoring of adherence have also been recommended by previous studies as an effective device for telecare monitoring of older adults and individuals diagnosed with unpredictable health conditions living independently. However, these devices should also be tailored for people with visual, hearing and dexterity problems [120]. A related issue shown in a previous qualitative study highlighted the importance of locking smart medication devices to prevent any interference by children [121], but it should still be easy to open for people with dexterity problems. However, all digital medication interventions should address privacy concerns [42,45,49,59,88,103,104]. Terms and conditions educating users on data protection and ways to protect their privacy, password protection of data, and an option to consent to the collection and use of their personal data should be included [64,108].

There was only 1 study included in the review that reported factors influencing large-scale implementation and scalability of a digital medication adherence intervention [35]. However, the workshop discussions identified several economic and policy factors that significantly hampered the large-scale implementation of the smart medication device that was provided to the patients at hospital discharge and domiciliary care recipients in the ICB region, United Kingdom. Examples of this included a recent decline in demand for adult domiciliary care services resulting in care providers’ fears of revenue loss, care providers’ fears of being held responsible for medication errors by care recipients [111], and care recipients’ fears of losing disability benefits due to reduction in their care [112]. As a higher value of financial incentives would have also limited the cost savings from this intervention, policy changes to make the use of the device the usual care for individuals requiring medication prompting visits from care providers should be considered right from the beginning of any person’s care plan.

Only 1 study [35] included in this review focused on re-establishing effective medication intake habits after hospital discharge; this was also the only study included in the review that assessed and reported factors affecting the large-scale implementation of the intervention. While the program theories derived from the workshop discussions filled this gap to an extent, future digital medication adherence intervention studies should explore and report the wider social, policy, and economic contextual factors that affect the scalability of the interventions. This is important as insights into factors affecting scalability of previous interventions can help in developing more effective implementation strategies for future interventions [122]. Further research is also needed to explore health system and policy changes needed to effectively implement medication adherence devices in the context of hospital discharge.

Lessons learned from the cocreation workshops were that it was essential for future studies to assess the wider social, economic, and policy factors that can potentially affect the scalability before starting the implementation of an intervention. It was also important to allocate enough time at the beginning of projects to develop effective relationships with key professionals in the community who can support the effective implementation of smart medication devices. Effective implementation strategies could be co-designed with key professionals who could support large-scale implementation of digital medication adherence interventions in the community at the beginning of the project [123].

Additionally, assessment of “richness” of the studies showed that details regarding the content and delivery methods of medication adherence interventions were lacking in some studies, which somewhat hindered the exploration of additional CMOs [9,29,30,35,43,44,49,53,56,58,62,69-72,75-77,80-84,86,88,92-94,96-98,100,103-106,108]. Thus, it is recommended that future medication adherence intervention studies describe the context, content, and delivery of the interventions in detail [123]. Moreover, the wide variation in the methods used in assessing adherence made it difficult for reviews to compare results among studies [83,106]. Some studies assessed adherence using only self-report measures, which are prone to bias and overestimation of adherence. There are drawbacks in using objective methods as well; for example, many digital tools can only assess when the packaging is opened and closed but do not measure if the medication has been ingested. Therefore, assessing adherence by using a combination of objective and self-report measures has been recommended for a comprehensive assessment of adherence [124,125].

A key strength of this study was its rigorous adherence to Medical Research Council guidelines [123] to uncover the contexts and mechanisms related to the efficacy or inefficacy of digital medication adherence interventions using realist synthesis. Additionally, cocreation workshops with diverse stakeholders of a project aiming to implement a digital medication adherence intervention led to the identification of CMOs of the intervention, which led to triangulation of the data. It also led to a better understanding of the implementation of such interventions in the community in the United Kingdom. Given the rapidly changing social policy and social care environment, a further strength was the use of the repeated workshops at approximately a 3-month interval with the same participants, which also gained insights as the implementation of the device evolved.

However, a limitation is that the CMOs developed from the workshop discussions could not be empirically verified and further refined using the cost-effectiveness and the effectiveness data of the device in improving medication adherence and quality of life because of the challenges in recruiting participants to use it. Nevertheless, one of the strengths of this study was that it highlighted that, in the future, when designing potential projects for interventions of this nature, it is important to assess the wider contexts and design some kind of flexibility into any project plans so that if wider contextual changes emerge, there is scope for the project to change as well.

This study focused on individuals who had adequate capacity to self-manage medication intake, and so papers which focused on medication adherence of people with severe mental illnesses, dementia, and substance use were excluded. Thus, the program theories derived from this review may not be applicable for medication adherence interventions for these populations. Only program theories associated with improving medication adherence among unintentionally nonadherent individuals were included; therefore, these program theories may not be generalizable for interventions designed for individuals who are intentionally nonadherent.

Digital interventions are commonly used to improve medication adherence; however, their efficacy may be limited because of a lack of effective tailoring for different populations and contexts [8]. Thus, this realist review addresses this gap in the literature by providing recommendations for tailoring digital medication adherence interventions effectively. Medication intake reminders and reminders for keeping water ready before medication intake, self-monitoring, and monitoring and feedback on medication adherence by device providers or health care professionals were the main active ingredients of interventions to reduce unintentional nonadherence across all theory areas. That said, intervention developers and implementers should adapt these active ingredients for specific contexts and patient subgroups by considering the recommendations provided in the CMO configurations identified in this study. This is important as the same active ingredients will not be effective across contexts because of how contextual factors or patients’ responses interact with the mechanism resources. For example, a person with visual impairment may be better supported by a device that provides auditory medication intake reminders (eg, Alexa) rather than pillboxes or mobile apps. Some CMOs in this paper suggest different delivery methods of medication adherence interventions in specific contexts; for example, CMO 11 suggests not including PRN medicines in medication devices and instead providing text message reminders or including written notes in the device as reminders to take them. In these situations, the intervention implementers should choose the delivery methods that align with the patients’ needs and preferences [123] and avoid implementing the mechanisms that can potentially result in unintended or negative consequences.

The review also provided insights into the economic, policy, and organizational factors affecting implementation and ways to overcome them. However, further studies exploring these factors are required, especially in the context of domiciliary care provision and hospital discharge. Future interventions should report intervention content and delivery in detail and codevelop effective implementation strategies with professionals who can support the large-scale implementation [123].